gemini-deep-research
Pass
Audited by Gen Agent Trust Hub on Apr 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill communicates with the official Google Generative Language API at
generativelanguage.googleapis.comto perform research tasks. This is a trusted service used for the skill's primary function. - [COMMAND_EXECUTION]: The skill includes and executes a Python script,
scripts/deep_research.py, which manages the research workflow, interacts with the Gemini API, and saves the resulting reports to the project workspace. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web data gathered during research, which is a known attack surface for search-based agents.
- Ingestion points: Untrusted content enters the agent's context through the Google API response in
scripts/deep_research.py. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the script's output handling.
- Capability inventory: The skill has access to
read_file,write_file, andrun_terminaltools, and the script itself performs local file writes. - Sanitization: The skill does not sanitize or filter the content retrieved from the web before generating markdown and JSON reports.
Audit Metadata